Refining the borderline personality disorder phenotype through finite mixture modeling: implications for classification.

Borderline personality disorder (BPD) is characterized by considerable heterogeneity. Prior approaches to resolving heterogeneity in BPD pathology have used factor and cluster analytic as well as latent class analysis strategies. These prior studies have been atheoretical in nature, but provide an initial empirical corpus for further sub-typing efforts in BPD. A model-based taxonomy for BPD that is supported by evidence from an advanced statistical methodology would enhance investigations of BPD etiology, pathophysiology, and treatment. This study applied finite mixture modeling analysis, in a model-guided fashion, to selected dimensions of pathology within a group of well-characterized BPD patients to determine if latent groups are harbored within the disorder. Subjects with BPD (N = 90) were examined on a variety of model-relevant psychopathology dimensions. We applied finite mixture modeling to these dimensions. We then evaluated the validity of the obtained solution by reference to a variety of external measures not included in the initial mixture modeling. Three phenotypically distinct groups reside within the overall BPD category. Group-1 is characterized by low levels of antisocial, paranoid, and aggressive features. Group-2 is characterized by elevated paranoid features, whereas Group-3 is characterized by elevated antisocial and aggressive features. External correlates reveal a pattern of differences consistent with the validity of this proposed grouping structure. A theory-guided finite mixture modeling analysis supports a parsing of the BPD category into three subgroups. This proposed BPD taxonomy represents an approach to reducing heterogeneity observed among BPD patients and it may prove useful in studies seeking to understand etiologic and pathophysiologic factors as well as treatment response in BPD.